Generally, an image sensor consists of a 2-dimensional array of a plurality of pixels. Each pixel outputs an electric signal corresponding to brightness of an input light. Thus, an image for each pixel may be formed using the electric signal.
In the manufacture of the image sensor, a procedural error may cause a fault pixel such as a dead pixel or a hot pixel, thereby deteriorating the quality of the image sensor. Pixel data generated by the fault pixel is excessively larger or smaller than pixel data generated by an adjacent pixel.
Techniques for removing the hot pixel are disclosed in, for example Korean Patent Publication Nos. 1020060042312 and 1020060038681.
The above-mentioned techniques combine N number of images taken with the reduced shutter time to minimize a hot pixel that may exist in the images. The techniques have disadvantages that 1) N number of frame buffer memories are required to store a plurality of images for combining the N number of images taken with the reduced shutter time, and 2) post-processing is required, for example image stabilization to remedy shaky hands at the time of shooting or compensation for color deterioration that may occur due to the reduced shutter time.
Meanwhile, an image processing apparatus generally requires a means for removing noise in an image inputted through an image sensor.
Techniques for removing noise are disclosed in, for example Korean Patent Publication Nos. 1020020093076 and 1020020093073.
The above-mentioned techniques conduct averaging by use of differences in level between an interest pixel and adjacent pixels, and a preset standard range. According to the techniques, in the case that the standard range is improper, noise in an image is not effectively removed. That is, the techniques have disadvantages that 1) an excessively large standard range blurs out a heterogeneous region of a dark portion in an image, and 2) an excessively small standard range fails a desired removal of noise in a light portion of an image.